memtest86

[TnT]

So I got more RAM for my computer now. I'm running a whole gig of RAM! The purchase was spawned by my data mining project that caused my computer to chug due to the excessive amount of data I was dealing with.

Anyways, I ran memtest86 for a while. On the first pass I got an error, but after 11 passes that was the only one. I've never had an error before, but it seems the error was an anomoly. Should everything be okay since I did 11 passes without a reoccurence or should I contact Crucial?

[KeiFront]

It probably just was a bit error that's why parity and ECC was invented. So probably the memory is fine, but it won't hurt to test the memory for a longer time if the error comes up again on the same address I would contact Crucial.

Anybody else any ideas?

[Fat Tone]

If you only surf and play games its maybe not a problem, but you shouldn't get a single error, ever. If your work matters, there's a problem. IMHO.

[Kurt]

++

[az]

Get new RAM. These tests can't be complete, they cover some cases where problems often arrive, and some random cases, but never ever can they cover EVERY state the RAM could be in.

Those very small errors can be more dangerous than the really big ones: When your memory is really faulty, you will know, because the machine may be unstable etc. But when you only have one bit error, you may well not notice because the machine runs perfectly stable, only that somewhere in your data, a bit is flipped. Maybe it's only the last bit so only one of your values is off by one, but maybe it's also the first bit on a large signed value, giving you -26657841 instead of 26657841. That's what ECC is for - catching those errors and alerting you.

AZ

[Wombat]

Any error is a bad thing. Run a whole lot more testing.

Also, you should be using memtest86+, not regular memtest86.

[Maggi]

disable all ECC features in your mainboard's BIOS and rerun memtest+

in case of any error, get the RAM replaced

really, there's no point in getting new hardware that is faulty from the beginning and the built in error correction is no excuse, but rather a life extension, if the hardware was fine directly after purchase.

[TnT]

Okay I've been running memtest all night (can't remember if it is plus or not, it's on the Gentoo LiveCD). After 16 runs no errors, but to be safe I'll contact Crucial for an exchange. I was so happy to finally be able to have everything I have open w/o hitting the swap space I didn't want to lose that for the short time it'll take to ship everything.

Update.
It is memtest86+. I ran it all night and day (86 passes I think, I forgot now), but no errors. The original error was in the 863 MB or something near that. Should I return both sticks or just the one in the 3rd slot? Mobo is dual channel. Does the number of megabytes correspond to where the RAM is located (ie 0-255 MB in slot a, 256-511in b, ...)?

[KeiFront]

Maybe it just was a single bit-error caused by some dust or electrical interference? So you just got lucky/unlucky to catch one . Does anyone know of the changes that this will/could happen?

[Maggi]

if you disabled all error correction in your BIOS prior to running memtest and the error didn't show up again, you should be safe to keep the RAM ...